cache-bypass explicitly reference pages involved in background i/o
authormatt@linuxbox.com <matt@linuxbox.com>
Fri, 18 Jun 2010 18:27:07 +0000 (14:27 -0400)
committerDerrick Brashear <shadow@dementia.org>
Fri, 2 Jul 2010 16:55:39 +0000 (09:55 -0700)
commitc66ba55bd882de677e425a9b9f7f29215a05b647
treeeb85c87b18bb3fd199c777982309002c2073444b
parent63304d6b2b895404e3a1a194232f3a48e448e75d
cache-bypass  explicitly reference pages involved in background i/o

Formerly, we assumed that any page eligible for background i/o could
be effectively pinned by lock_page.  Persuant to concern expressed by
Simon that such pages might be eligible to be released by the kernel
after readpages returns, call get_page to increment the refcount on
each page before dispatching a background read (and matching put_page
when the i/o is completed).

Change-Id: Ib3a63e56e6b902b4eb5deb769847e7f17ce2c9ff
Reviewed-on: http://gerrit.openafs.org/2215
Reviewed-by: Derrick Brashear <shadow@dementia.org>
Tested-by: Derrick Brashear <shadow@dementia.org>
src/afs/LINUX/osi_vnodeops.c
src/afs/afs_bypasscache.c